Brain and behavioral differences in speech segmentation between typically developing children and children with ASD

Wagley, N. ¹ , S.F. Hay, J. ² , Ugolini, M. . ¹ , M. Bowyer, S. ³ , Lajiness-O'Neill, R. . ⁴ , Brennan, J. ⁵ & Kovelman, I. ¹

1 Dept. of Psychology, University of Michigan, USA
2 Dept. of Psychology, University of Tennessee, USA
3 Dept. of Neurology, Henry Ford Hospital, Detroit USA
4 Dept. of Psychology, Eastern Michigan University, USA
5 Dept. of Linguistics, University of Michigan, USA

Speech segmentation, or identifying syllables and words in a continuous linguistic stream, is foundational for language learning. Autism Spectrum Disorders (ASD) is typically characterized by impairments and delays in language acquisition. Here we explored the hypothesis that, compared to neurotypical peers, children with ASD may have difficulty extracting statistical regularities from a naturalistic linguistic stream. English-speaking children with ASD (ages 8-12, n=15) and healthy controls (n=14) listened to three repetitions of a 2 minute passage in a new language (Italian). The children?s brain activity was measured with magnetoencephalography (MEG). The passages were produced in a child-directed manner and included two target words with high (TP= 1.0; HTP) and two words with low (TP= 0.33; LTP) internal predictability between syllables. Behavioral testing at the end of the session revealed that typically developing children (68% correct) were better at learning the new HTP and LTP words than the children with ASD (41% correct; t(24) = 4.18, p < 0.001). Preliminary MEG measures of neural coherence and word-level event related responses also suggest differences between the groups. The findings contribute new insights to better understanding of how the developing brain segments the linguistic stream and the etiology of language impairments in ASD.